Cooperativity of papain-substrate interaction energies in the S2 to S2

Feb 5, 1991 - Herschel Wade and Thomas S. Scanlan. Journal of the ... John D. Doran, Peter J. Tonge, John S. Mort, and Paul R. Carey. Biochemistry 199...
0 downloads 0 Views 1MB Size
Biochemistry 1991, 30, 1394-1402

1394

and if so the effect of this on kinetic measurements. Mitochondrial forms of the enzyme have been shown to increase in activity upon dissociation (Dickinson & Allanson, 1985; Weiner & Takahashi, 1985). Registry No. NADH, 58-68-4; propionaldehyde, 123-38-6; aldehyde dehydrogenase, 9028-86-8.

REFERENCES Bennett, A. F., Buckley, P. D., & Blackwell, L. F. ( 983) Biochemistry 22, 776-784. Blackwell, L. F., Bennett, A. F., & Buckley, P. D. ( 983) Biochemistry 22, 3784-3791. Blackwell, L. -F., Motion, R . L., MacGibbon, A. K. H., Hardman, M. J., & Buckley, P. D. (1 987) Biochem. J . 242, 803-808. Buckley, P. D., & Dunn, M. F. ( 1 982) Prog. Clin. Biol. Res. 114, 23-35. Dawson, R. M . C., Elliot, D. C., Elliot, W. H., & Jones, K. M. ( I 969) Data for Biochemical Research, pp 483-500, Oxford University Press, New York. Dickinson, F. M. ( 1 985) Biochem. J . 225, 159-1 65. Dickinson, F. M. ( 1 986) Biochem. J . 238, 75-82. Dickinson, F. M., & Hart, G. J. (1982) Biochem. J . 205, 443-448. Dickinson, F. M., & Allanson, S. (1 985) Prog. Clin. Biol. Res. 174, 71-82.

Dickinson, F. M., & Haywood, G. W. (1986) Biochem. J. 233, 877-88 3 . Dickinson, F. M., Hart, G. J., & Kitson, T. M. (1981) Biochem. J . 199, 573-579. D u m , M. F., & Buckley, P. D. (1 985) Prog. Clin. Biol. Res. 174, 15-27. Hart, G . J., & Dickinson, F. M. (1982) Biochem. J . 203, 61 7-627. Hempel, J . , & Jornvall, H. (1 987) Prog. Clin. Biol. Res. 232, 1-14. Klotz, I. M., Darnall, D. W., & Langerman, N. R . (1975) Proteins (3rd Ed.) I , 293-411. MacGibbon, A. K. H., Buckley, P. D., & Blackwell, L. F. (1977a) Biochem. J . 165, 455-462. MacGibbon, A. K. H., Blackwell, L. F., & Buckley, P. D. (1977b) Biochem. J . 167, 469-477. MacGibbon, A. K. H., Blackwell, L. F., & Buckley, P. D. ( 1 9 7 7 ~ )Eur. J . Biochem. 77, 93-100. MacGibbon, A. K. H., Motion, R. L., Crow, K. E., Buckley, P. D., & Blackwell, L. F. (1979) Eur. J . Biochem. 96, 58 5-595. Segel, I. H. (1 975) Enzyme Kinetics. Behavior and Analysis of Rapid Equilibrium and Steady-State Enzyme Systems, pp 506-523, Wiley, New York. Weiner, H., & Takahashi, K. (1985) Prog. Clin. Biol. Res. 174, 83-90.

Cooperativity of Papain-Substrate Interaction Energies in the S2 to S i Subsitest Paul J. Berti,t.g Carlos H. Faerman,: and Andrew C. Starer*.* Protein Engineering Section, Biotechnology Research Institute, 6100 avenue Royalmount, MontrZal, QuPbec, H4P 2R2, Canada, and Department of Biochemistry, McGill University, 3655 Drummond Street, MontrPaI, QuPbec, H3G I Y6. Canada Received August 21, 1990; Revised Manuscript Received October 17, 1990

ABSTRACT: Enzyme-substrate contacts in the hydrolysis of ester substrates by the cysteine protease papain were investigated by systematically altering backbone hydrogen-bonding and side-chain hydrophobic contacts in the substrate and determining each substrate's kinetic constants. The observed specificity energies [defined as AAGOb = -RT In [(k,,t/Kar)mt/(k,,~/K~)~~"~)]] of the substrate backbone hydrogen bonds were -2.7 kcal/mol for the P2 NH and -2.6 kcal/mol for the PI NH when compared against substrates containing esters a t those sites. T h e observed binding energies were -4.0 kcal/mol for the P2 Phe side chain, -1.0 kcal/mol for the PI' C=O, and -2.3 kcal/mol for the P2'NH. T h e latter three values probably all significantly underestimate the incremental binding energies. The P2 NH, P2 Phe side-chain, and PINH contacts display a strong interdependence, or cooperativity, of interaction energies that is characteristic of enzyme-substrate interactions. This interdependence arises largely from the entropic cost of forming the enzyme-substrate transition state. As favorable contacts are added successively to a substrate, the entropic penalty associated with each decreases and the free energy expressed approaches the incremental interaction energy. This is the first report of a graded cooperative effect. Elucidation of favorable enzyme-substrate contacts remote from the catalytic site will assist in the design of highly specific cysteine protease inhibitors.

P a p a i n (EC 3.4.22.3) is the most studied and consequently best understood member of the papain superfamily of cysteine proteases. The hydrolytic mechanism is minimally represented kinetically by a three-step process: 'NRCC Publication No. 32396. *Biotechnology Research Institute. 5 McGill University.

0006-2960/91/0430-1394$02.50/0

E

+S& E 6 5EA + Prodl k-,

-+ k+I

E

Prod2

The free enzyme and substrate (E + s) associate to form a Michaelis complex ( E 4 ) with a dissociation constant Ks (= l ~ - ~ / l C + The ~ ) . carbonyl carbon of the scissile ester or amide bond undergoes nucleophilic attack from the active-site cysteine (Cys 25), forming the covalent acyl-enzyme intermediate (EA) and releasing the alcohol or amine product (Prodl) with 0 1991 American Chemical Society

Papain-Substrate Interaction Energies a rate constant k+2. The acyl-enzyme then deacylates with a rate constant k+3 by undergoing nucleophilic attack from a water molecule to yield the free enzyme and the carboxylic acid product (Prod2). The acylation and deacylation steps are believed to proceed through tetrahedral intermediates [e.g., see Brocklehurst et al. (1987)l. Enzymic Catalysis. Enzymes catalyze reactions by preferentially stabilizing the transition state of the reaction over the reactants’ ground state. Transition-state stabilization by enzymes can be achieved by two means, by using enzyme/ Substrate binding energy to effect ground-state destabilization (Jencks, 1975), thereby lowering the free energy of transition-state formation, and by compensation for entropy loss. Entropy is lost on forming an enzyme-substrate transition state due to the loss of translational and overall rotational motion of the two molecules relative to each other. Entropy loss on formation of the transition state is only weakly dependent on the identity of the substrate and has been variously estimated to be on the order of -35 to -50 cal/(Kmol) for reactions with “tight” transition states, including acyl transfer reactions (Page & Jencks, 197 1; Finklestein & Janin, 1989), corresponding to 10-15 kcal/mol at 300 K. If an enzyme does nothing more in catalyzing a bimolecular reaction than to bind the two reactants in the active site, then the free energy of activation will be reduced by =l0-15 kcal/mol.’ lncremental BindinglSpecficity Energies. In examining a particular functionality of a substrate, one seeks to determine its incremental binding energy. As described by Fersht (1988), several criteria must be met in order for an observed binding energy of a substituent to be considered an incremental binding energy. It is necessary to delete the substituent under investigation, since substitution with another group that can interact with the enzyme will lead to a strict specificity effect; the deletion must be such that solvent has free access to the binding site on the enzyme, otherwise the free energy cost of excluding solvent will be included in the observed energies; and finally, there must be no difference in the entropy of the uncomplexed substrates and in the enzyme-substrate complexes in the presence and absence of the substituent. The latter condition becomes only approximately true for very good substrates; so one may only determine, at best, the lower limit of a group’s incremental binding energy. The observed binding energies, AAGob, are determined from

AAGobs = -RT In [(kcat/KM)A/(k~at/~M)B] where k,,/KM is the specificity constant for substrates A and B. k,,/KM, which equals k+z/Kswhen k-, >> kf2, is inversely proportional to the dissociation constant for the quasi-equilibrium between the free enzyme plus free substrate to the enzyme-substrate transition state. There is a difficulty in determining the incremental binding energy of a functionality such as a substrate’s P I NH in papain-catalyzed ester hydrolysis insofar as that the requirement of no entropy difference between different substrates requires the use of a substrate that extends into at least the S2subsite to achieve a sufficiently high k,,/KM. Obviously, deletion of the P, N H is then impossible. If that functionality is substituted with another, then potential interactions of the new substituent with the enzyme will interfere with determination of the incremental binding energy; that is, in this instance AAGob, is purely a specificity energy rather than a binding energy. When considering a substitution, one can only deI We neglect here the entropy changes involved in releasing bound water from thc cnzymc and ligand and changes in the vibrational entropies on complex formation.

Biochemistry, Vol. 30, No. 5, 1991

1395

termine a quantity that will be referred to here as the incremental specifcity energy, by analogy to the incremental binding energy. Use of the phrase incremental specificity energy makes the same assumptions as determination of incremental binding energy except that the effect of the interaction of the enzyme with the new substituent is taken into consideration. If the free energy of interaction of the new substituent can be determined independently, then the incremental binding energy can be estimated indirectly. . Specficity. Early studies on the specificity of papain included the work of Schechter and Berger [ 1967, 1968; see also Berger and Schechter (1970)l who demonstrated seven subsites in papain’s active site, four on the acyl group side (S4-Sl)and three on the leaving group side (S1’-S3’).2 They identified a specificity for Phe at P2. The specificities of a number of other cysteine proteases have also been investigated [e.g., see Brocklehurst et al. (1987)l. The study of cleavage patterns of a number of polypeptides has failed to yield clear specificity requirements for any of the cysteine proteases examined. The inability to discern unambiguous specificites at any single subsite from peptide cleavage maps demonstrates that a number of enzyme-substrate contacts at different locations contribute to specificity and that no single contact is indispensable [cf. serine proteases that have very strong specificities at P I , such as trypsin, which requires Lys or Arg, or Staphylococcus aureus protease V-8, which requires Asp or Glu at PI (Polgar, 1987)l. Because of the implication of cysteine proteases in a number of disease states, including myocardial infarction, cancer metastasis, muscular dystrophy [see Rich (1 986)], and parasitic (Dresden et al., 1982; Coombs, 1982) and bacterial pathogenicity (Bjorck et al., 1989; Shah & Gharbia, 1989), recent specificity studies have focused on the inhibition of cysteine proteases, or more specifically, on selectivity inhibiting one cysteine protease in the presence of others [see Demuth (199O)l. Selective inhibition of particular cysteine proteases in vivo would also allow elucidation of the roles of those proteases. Studies on the specificity of inhibition have used side-chain variation of peptide-analogue inhibitors to effect specificity for a particular protease. The contribution of the peptide backbone to specificity has received less attention (Asboth et al., 1988). In this study we have identified and characterized hydrogen-bonding interactions between peptide substrate backbone atoms in the S2-S; subsites as well as side-chain interactions in the S2subsite. We discuss the importance of using these hydrogen bonds in maximizing the differentiation between the reactivities of different cysteine proteases. While the importance of compensation for entropy loss has been well recognized in theoretical discussions (Page & Jencks, 1971; Jencks, 1975, 1981), to our knowledge this is the first exp.erimental report of a graded effect. The observed binding or specificity energies for a given substituent are dependent strongly on the presence of other enzyme-substrate interactions; that is, they show a strong cooperativity. MATERIALS AND METHODS

Materials Mercuripapain was prepared from papain (Sigma Chemical Co., Type 111) by the method of Sluyterman and Wijdenes (1970) and activated as described previously (Storer et al., 1983). Freshly activated papain contained 1.O mol of SH/mol Subsite numbering proceeds outward from the catalytic site. The corresponding amino acid residues in the peptide substrate are numbered: P,-P,-P,-P, tP,’-P,’-P,’, where f indicates the scissile amide or ester bond.

1396 Biochemistry, Vol. 30, No. 5, 1991

Berti et al.

overnight, with warming to room temperature. The solution of protein and was stored on ice until used. was rotary-evaporated to dryness, then ca. 30 mL of CHzC12 H PLC3-grade acetonitrile and high-purity H 2 0 (resistivity was added. The precipitate was filtered off after sitting ov> I8 MQcm) were used for HPLC. All other reagents and solvents were the highest available grade. ernight, and the material was purified by silica gel flash chromatography in CH,CI2/methanol (95:5, v/v). 14 was All compounds gave acceptable elemental analyses further purified by HPLC in (0.1% TFA/H20)/acetonitrile [fO.O2(calculated percentage), except %N, which was (calculated percentage) f 0.2%] and gave only one peak by (55:45, v/v). 15 was purified by HPLC in (0.1% TFA/ HPLC. Analytical HPLC was performed on C-18, end-capH20)/acetonitrile/H20 [times are cumulative: 0 min, 0:5050 ped, reverse-phase columns, 4 mm X 15 cm with 5-pm ir(v/v/v); 4 min, 50:50:0;28 min, 50:50:0;38 min, 20:80:0;flow regular particles, with absorbance detection at 205 nm. = 5.0 mL/min]. Both were white solids (yield = 240 mg for Preparative HPLC used a C- 18 reverse-phase column, 19 mm 14, 150 mg for 15). Anal. Calcd for 14: C, 60.19; H,6.13; X 30 cm, with absorbance detection at 205 nm. N, 5.04. Found: C, 59.89; H, 6.11; N, 5.20. Anal. Calcd AcPheGlyOMe (1). AcPhe [synthesized as described prefor 15: C, 61.41; H, 6.53; N, 4.79. Found: C, 61.21; H, 6.28; N, 4.96. viously (Carey et al., 1984)] and GlyOMe were coupled by the mixed anhydride method as described previously (Carey O-Acetyl-/3-phenyllactylGlyOMe (2). L-/3-phenyllacticacid et al., 1984). Anal. Calcd for 1: C, 60.42; H, 6.52; N, 10.07. was synthesized from L-phenylalanine by treatment with Found: C, 60.59; H, 6.59; N, 9.89. nitrous acid (Cohen & Weinstein, 1964). O-acetyl-L-0Glycolic Acid Methyl Esters and Glycolamides [ MocPhephenyllactic acid was synthesized by refluxing with acetic anhydride (Ingles & Knowles, 1968). 2 was then made by O-CH,-C(0)-OMe (4),Ph-(CH2),-C(0)-O-CH,-C(O)-OMe (8), MocPhe-O-CH,-C(0)-NH, (13), and Ph-(CH,),-Cthe mixed anhydride method as used in Carey et al. ( 1 984) (0)-O-CH,-C(0)-NH, (17)]. MocPhe (6;Carey et al., 1984) and purified by silica gel chromatography in CH2C12/methanol or Ph-(CH,),-COOH (10)(1.1 equiv, approximately 10 mmol) (9:1 , v/v). The material was rotary-evaporated to give a yellow was dissolved in ca. 25 mL of acetone with 1 equiv of methyl oil. The compound was characterized by NMR: 6 ([,H6]bromoacetate or iodoacetamide and 1 equiv of anhydrous acetone) 2.1 [3 H, s, CS3-C(O)], 3.1-3.4 (2 H, m, CHK2C03,and the mixture was refluxed for 5 h. The acetone CH,-Ph), 3.8 (3 H, s, C(O)-O-CH,), 4.0 [2 H, d, J = 7 Hz, was removed under vacuum. Remaining material was disNH-CH,-C(O)], 5.4 [ l H, m, 0-CH-C(O)], 7.4 (5 H, s, solved in ca. 30 mL of water and ca. 30 mL of CH,C12. The c&5), 7.7 ( 1 H, br s, NH). Stoichiometry of the hydrolysis organic layer was washed with 1 N HCI, 1 N NaOH, and by papain showed 97.8% purity. saturated NaCl solution, placed overnight over Na,SO,, and MocPheGly-OMe (3;Angus et al., 1986) and N-benzoylthen filtered, and the solvent was removed under vacuum. The Gly-OMe (18;Storer & Carey, 1985) have been described remaining material was dissolved in methanol, filtered, and previously. MocGlyGly-OMe (ll),MocPhe-OMe (16),and dried under an N, stream, then extensively under vacuum. benzoyl-O-CH,-C(0)-OMe (19)were the generous gifts of Anal. Calcd for 4: C, 56.95; H, 5.80; N, 4.74. Found: C, Dr. Ron Angus. 57.09; H, 5.79; N, 4.84. Anal. Calcd for 8: C, 64.85; H, 6.35. Methods Found: C, 64.57; H, 6.44. Anal. Calcd for 13: C, 55.69; pH-stat kinetics were performed as previously described H, 5.75; N, 10.03. Found: C, 55.19; H, 5.81; N, 10.20. Anal. (Storer & Carey, 1985). Reaction solution was 300 mM Calcd for 17: C, 63.74; H, 6.32; N , 6.79. Found: C, 63.47; NaCI, 1 mM EDTA, and 20% (v/v) acetonitrile, pH 6.0, with H, 6.34; N, 6.65. enzyme concentrations of 0.14-6 pM. Kinetics constants were MocPhe- 0-CH2-COOH (5) and Ph-( CH,) ,-C( 0)0determined directly from initial velocities ( o ~ / [ E ] ~by) the CH2-COOH (9). 13 and 17 were deamidated by the method algorithm of Cornish-Bowden (1976), which assumes that of Nefkens and Nivard (1965). 5 was purified by flash random errors in observed u,/[E],’s are proportional to the chromatography on silica gel in ethyl acetate/hexane (7:3, v/v) true u,/[E],; this is true for pH-stat kinetic determinations and dried under an N z stream, then under vacuum. HPLC (Storer & Carey, 1985). analysis indicated a ratio of [5]/[6] of 32:l. It was further MocPhe- 0-CH,- C( 0)- OMe (4) and Ph-( CH,) C(0)- 0purified by preparative HPLC in (0.1% TFA/H20)/acetoCH,-C(0)-OMe (8). These two substrates each contain two nitrile (7:3, v/v) to give a ratio of 500:l. The pure resinous ester functionalities. In each case, hydrolysis may proceed at material was dried under vacuum and characterized by NMR: either or both esters; for example, at MocPhetO-CH,-C6 ([2H,]methanol) 2.91-2.96, 3.25-3.30 (2 H, 2 m, CH(0)-OMe and MocPhe-O-CH,-C(0) tOMe (where t indicates CH,-Ph), 3.53 ( 3 H, s, O-CH,), 4.51-4.54 [ I H, m, NHthe scissile bond). If it is assumed that productive binding for CH-C(O)], 4.62-4.72 (2 H, m, O-CH,-COOH), 7.22 ( 5 H, hydrolysis of one ester functionality precludes hydrolysis at m, C6H,). Anal. Calcd: C, 56.95; H, 5.80; N, 4.74. Found: the other without the Michaelis complex first dissociating and C, 57.09; H, 5.79; N, 4.84. 9 was purified by HPLC in (0.1% reforming (i.e., effectively binding a different substrate), then TFA/H20)/acetonitrile (4:6 v/v). giving white crystals. Anal. hydrolysis at each ester functionality proceeds as it would for Calcd: C, 63.45; H, 5.81. Found: C, 63.21; H, 5.86. two chemically distinct competitive substrates (Berti, 1990). MocPhe-O-CH2-C(0)-CH, (14) and MocPhe-O-CH,-CIn this case (0)-CH2CH3(15). 6 (5 mmol) was dissolved with stirring in tetrahydrofuran with 6 mmol of I-hydroxypropanone or 1(kcat/KM)T = ( k c a t / K M ) B + (kcat/KM)C hydroxy-2-butanone, plus 5.5 mmol of 1 -hydroxybenzotriazole and and 3 drops of N-methylmorpholine at -10 OC; then 5.5 mmol of dicyclohexylcarbodiimide was added. Stirring was continued O0.B /OO,C = ( k a t / K M ) B / ( k c a t / KM)C

,-

Abbreviations: All amino acids and derivatives are abbreviated as recommended by the IUPAC-IUB Joint Commission on Biochemical Nomenclature ( 1 985); HPLC, high-pressure liquid chromatography; Moc-, (methoxycarbony1)-; TFA, trifluoroacetic acid; t, in substrate names, the scissile ester bond.

where T refers to the total reaction and B and C refer to hydrolysis at the individual ester functionalities. To determine the kinetic constants for hydrolysis at each site, it was necessary to separate and quantitate the reaction products by HPLC. Analytical HPLC was performed as

Papain-Substrate Interaction Energies described above, with flow = 2.5 mL/min for all runs. For 4 and its products, elution was with (0.1% TFA/H,O)/ acetonitrile, from 9:l (v/v) to 3:7 over 6 min. For 8, the same solvents were used, from 75:25 to 35:65 over 6 min. Standard curves were determined for both substrates and each of the potential products: 5, 6, 9, and 10. In quantitating substrate and products in the reaction mixtures, it was necessary to quickly and gently separate them from papain, which irreversibly bound to and destroyed the HPLC column. This was accomplished by a dialysis arrangement using two nested test tubes separated by dialysis tubing with a cylindrical diameter of 1.6 cm (2.0 mL/cm) and a molecular weight cutoff of 3500. The outer test tube was a silanized 30-mL Corex centrifuge tube and contained 20 mL of reaction mixture. The inner test tube was a 13 mm X 100 mm disposable culture tube. Use of the inner test tube considerably increased the surface area to volume ratio and therefore the rate of dialysis. A small volume (200 pL) of reaction buffer was placed inside the dialysis membrane and allowed to equilibrate for 5-10 min before being removed and injected directly onto the HPLC column. The reaction buffer was 100 mM sodium phosphate, 200 mM NaC1, pH 6.0 and 20% (v/v) acetonitrile. This is the same ionic strength, percentage acetonitrile, and pH as the pH-stat reaction solution. Phosphate ions have been shown not to have an effect on papain activity (Whitaker & Bender, 1965; Storer & Carey, 1985). Samples taken at various times from the dialysis setup in the reaction of papain with substrate were directly injected onto the column. For the reactions with 4,the resulting peak areas for each component were converted to nanomoles injected, and the initial substrate concentration was then used to calculate the concentration of each component present in the reaction mixture from the relative amounts present. The ratio of products [5]/[6] was also determined. Initial velocities for substrate depletion and product formation were determined by the direct linear plot method (Cornish-Bowden, 1975). Michaelis-Menten kinetic constants determined from HPLC reactions were identical within experimental error with those determined by pH-stat (not shown). The products of this reaction that still contained an ester functionality were examined as substrates. Glycolic acid methyl ester does not produce a measurable rate by pH-stat when 2 mM is reacted with 2 pM papain. 5 was tested as substrate by reacting 3.3 mM with 15 pM papain for 60 min in the dialysis setup and then submitting the products to HPLC analysis with the solvent program as for 4. MocPhe-0-CH2-C(0)-NH2(13). Compound 13 was reacted with papain by the dialysis method to confirm hydrolysis at the ester. 13 from the dialysis setup before addition of papain gave a single peak. After a 15-min reaction of 1.9 mM 13 with 2 pM papain, a gradient of (0.1% T F A / H 2 0 ) / acetonitrile [ 1OO:O (v/v) to 35:65 over 6 min at 2.5 mL/min) gave peaks for 13 and 6. Ph-(CH2),-C(0)-0-CH2-C(0)-~Hz (17). As no reaction rate could be detected by pH-stat, 17 was tested as a substrate by HPLC. 17 in the dialysis setup before addition of papain gave a single peak with (0.1% TFA/H,O)/acetonitrile [8:2 (v/v) to 35:65 over 6.5 min at 2.5 mL/min]. After reacting 5 mM 17 with 2 pM papain for 80 min, only 17 was found by HPLC. Spiking an aliquot of the 80-min sample with Ph-(CH,),-COOH gave two peaks.

RESULTS MocPhe- 0CH2-C(0)-0 M e (4).Initial velocities (uo/ [E]0)s) were calculated for 4 depletion and for 5 and 6 formation. In

Biochemistry, Vol. 30, No. 5, 1991

1397

all cases they appear to follow Michaelis-Menten kinetics, as would be expected. The u,/[E],’s calculated for 4 depletion were identical within experimental error with those determined by pH-stat. The ratio of products formed, [5]/[6], was 8.2 f 0.4. Within experimental error, the ratio [5]/[6] was independent of elapsed time or [SI, (not shown). On the basis of the HPLC runs of 4 reacted with papain, particularly in light of the constant [5]/[6] with time, 5 would be expected to be an extremely poor substrate, if at all. After incubating 3.3 mM 5 with 15 pM papain for 60 min, 0.186 mM 6 had formed, giving u0/[E], = 3.43 X s-l. By using the Michaelis-Menten equation and assuming k,,, L 0.1 S-I, 5-fold less than the lowest k,, measured in this study, KM 1 92.9 mM is obtained, yielding kcat/KMI1.1 M-l-s-’. Thus 5 is a poor substrate, with an upper limit on the kcal/KMof 1.1 M-1.s-I. Ph-( CH,),-C( O)-O-CH,-C( 0 ) - 0 M e (8). 8 was examined by HPLC, like 4,to determine the products of the reaction with papain and their ratios of formation. 9 was detected and quantified, but 10 was not found. After a 76-min reaction time with 2 mM substrate and 6.2 pM papain, no 10 peak was detected. At 76 min, the 9 peak was 35.32 nmol. Peaks at least as low as 0.1 1 nmol 10 would have been detected. This gives a ratio [9]/[10] L 321. Given k,,/KM = 79 M-’.s-’ for 8a (see Table I), kcat/KMI0.24 M-’.s-] for 8b. Ph-(CHz),-C(0)-0-CH2-C(0)-NH2 (17). As no reaction rate could be detected by pH-stat, 17 was tested as a substrate by HPLC. By using the same arguments as with 8b and assuming the same absorptivity as 10, the 17 peak was 80.2 nmol. Therefore [lo] I0.0131 mM. This corresponds to uo/[ElOI1.36 X s-l. Assuming again that k,,, 2 0.1 s-I, then KM = 31.8 mM and kCa,/KM5 0.31 M-l-s-’. The results of the pH-stat and HPLC kinetics are summarized in Table I.

DISCUSSION By systematically varying substituents of substrates to fulfill specificity requirements for papain, we have been able to map out interactions in the S2-S,’ subsites. The free energies of the hydrogen bonding interactions at the P, NH, PI N H , PI’ C=O, and P,’ N H , as well as of hydrophobic interactions of the P2 Phe side chain, were characterized. Putative hydrogen-bonding partners were identified by reference to the X-ray crystallographic structures of a-chloroketone and stefin B inhibited papains (Drenth et al., 1976; Stubbs et al., 1990) (Figure 1). PI’ and P,’ Subsites. The kinetic constants of 13 [toCHz-C(O)-NHzI4and related substrates demonstrate that there are hydrogen-bonding partners for the Pl’-P,’ amide group of peptide substrates. This is the first report of these interactions. AAG,, = -2.2 kcal/moP between 13 [tO-CH,-C(O)-NH,] and 4b [tO-CH,-C(O)-OMe] could be attributed to hydrogen bonding or to unfavorable steric interactions between the ester’s methyl group of 4b [tO-CH,-C(O)-OMe] and the enzyme. Isosteric analogues of each compound that are not hydrogen bonding in this region were synthesized. The difference between 14 [ fO-CH,-C(O)-CH,] and 15 [tO-CH,-C(O)In this section, to facilitate the discussion of substrates involving the S,’ and S,’ subsites, the structure of the leaving group is indicated in

brackets after the compound number. Thus,MocPhetOCH2-C(0)-OMe is 4b [tOCH2-C(0)-OMe]. In all the schemes under Discussion, AAGob, = -RT In [ ( k a t / KM)n,,/(kcal/KM)sccond].A positive AAGob, means that the second substrate listed has a higher k,,/KM; a negative AAGob, means that the first substrate is better. Random errors in PACob, = f O . l kcal/mol.

1398 Biochemistry, Vol. 30, No. 5, 1991

Berti et al.

Table I: Kinetic Constants As Determined by pH-Stat and HPLC substrate k,,/KM (M-'*S-') 120000 f 5000 1 AcPheGlytOMe 2 O-acetyl-P-phenyllactyl-GlytOMe I100 100 39000 f 4000 3 MocPhcGlytOMe 4 MocPhc-0-CH2-C(0)-OMe (overall) 480 f 30 4a MocPhc-O-CH2-C(0)tOMeR 430 f 35 4b MocPhctO-CH2-C(0)-OMeR 52 f 4 Il.lb 5 MocPhctO-CH2-COOH 7 Ph-(CH2)2-C(0)-GlytOMec 710 f 30 8a Ph-(CH2)2-C(0)-O-CH2-C(O)~OMe 79 f 4 8b Ph-(CH2),-C(O)tO-CH2-C(0)-OMe 50.24b 40 f 2 11 MocGlyGlytOMe 12 AcGlytOEtd 2.4 f 0.2 13 MocPhetO-CH2-C(O)-NH2 2200 f 100 14 M o c P ~ c ~ O - C H ~ - C ( O ) - C H ~ 39 f 3 15 M o c P ~ c ~ O - C H ~ - C ( O ) - C H ~ C H ~ 96 f 5 16 MocPhctOMe 7.12 f 0.04 17 Ph-(CH2)2-C(O)tO-CH2-C(O)-NH2 5 0 . 31 18 N-benzoyl-GlytOMe I40 f 6 19 bcnzoyl-O-CH2-C(0)tOMe 17 f 2

*

k,, (s-9 9.4 f 0.2 2.1 f 0.2 6.3 f 0.2 12f5

4.7 f 0.2 0.67 f 0.07 1.8 f 0.2 1.4 f 0.1 4.7 f 0.2 1.6 f 1.0 15 f 28 e 2.8 f 0.2 0.23 f 0.04

K M (mM) 0.078 f 0.004 I .9 f 0.3 0.16 f 0.02 26f 11

no. of points 37 32 34 44

5.9 8.4 f 1.2

26

45 f 6 590 f 100 2.2 f 0.2 40 f 28 I60 f 300 e

22 16 48 29 27 9

20 f 2 14f4

33 49

'Calculatcd from the k c a l / K Mof the overall rate of hydrolysis of MocPhe-0-CH,-C(O)-OMe and the ratio of product formation: [5]/[6] = 8.2

f 0.4. bSee text for explanation. 'From Storer et al. (1988). dWilliams et al. (1972) found k,,,/KM = 2.25 M - ' d for AcGlytOMe as opposed to

2.4 M - k I for AcGlytOEt reported here, showing a negligible contribution from the ethyl versus the methyl leaving group and indicating that the comparison with Ph-(CH2),-C(O)-Glyt0Me is valid. KM >>

S2 Hydrophobic

Asp 158 C=O

Gly 66C=O,N H .

I

s3

~

s2

I

si

Si'

j

S2'

I

FIGURE I : Schematic representation of binding of a hypothetical in papain's active site substrate, N-acetyl-PheGlytO-CH2-C(0)-NH2, as deduced from the X-ray crystallographic structures of e-chloroketone-inhibited papains (Drenth et al., 1976) and the stefin Bpapain complex (Stubbs et al., 1990). The active-site cysteine (Cys 25), the S2hydrophobic pocket (Tyr 67, Pro 68, Val 133, Val 157, and Ala 160), and those residues hydrogen-bondingto the substrate (Gly 66 and Asp 158) are shown. The hydrogen-bonding partners on the leaving-group side (subsites SI' and S i ) are not known.

CH2-CH3] should reflect steric effects only. AAGobs = 0.5 kcal/mol indicates that there is not a problem sterically with the size of the 4b [tO-CH,-C(O)-OMe] leaving group and therefore that the difference between 13 [ tO-CH,-C(O)-NH,] and 4b [ tO-CH,-C(O)-OMe] is a hydrogen-bonding effect. The k,,/KM of 16 [tOMe] versus 13 [tO-CH,-C(O)-NH,] gives AAGob, = 3.3 kcal/mol. It is possible to separate the contributions from the C=O and -NH2. The hydrogenbonding contribution of the carbonyl portion may be estimated thus: Ahcobs = -1.0 kcal/mol between 4b [tO-CH,-C(O)OMe] and 16 [tOMe]. Lowe and Williams (1965) found that the k,,,/KM's of N-benzoyl-GlytOMe (133 M-l-s-'), f O E t (kcat/KM = 133 M-'*s-'), and t o - n P r (118 M-'*s-', AAGobs > -0.1 kcal/mol) were the same. Thus, the favorable AAGob between 16 [fOMe] and 14 [tO-CH,-C(O)-CH,] may be attributed to the binding of the carbonyl oxygen, with no apparent contribution from the methyl group or carbonyl carbon. For this comparison of N-benzoyl-GlytO-nPr with 14 [tO-CH,-C(O)-CH,] to be valid, it is necessary that the leaving-group carbons interact with the enzyme in the same conformation in each substrate. This assumption is supported

by noting that, on adding a methylene to either leaving group, the kcat/KMis improved by the same amount; Le., 14 [toCH,-C(O)-CH,] versus 15 [ tO-CH,-C(O)-CH,-CH,], AAGobs = 0.5 kcal/mol and N-benzoyl-GlytO-nPr versus N-benzoyl-GlytO-nBu, AAGob, = 0.4 kcal/mol. The AAGobs = -1 .O kcal/mol for the PI' carbonyl is therefore due primarily to the carbonyl oxygen forming a hydrogen bond with a donor on the enzyme. Given the contribution of -1.0 kcal/mol by the carbonyl group, the remaining -2.3 kcal/mol of the AAGobsbetween 16 [ tOMe] and 13 [ tO-CH,-C(O)-NH,] is contributed by hydrogen bonding of the -NH,. This is the same as the free energy observed when comparing 13 [ tO-CH,-C(O)-NH,] with 14 [ ?O-CH,-C(O)-CH,]. Given the unusual situation here that the terminal methyl of 14 [tO-CH,-C(O)-CH,] does not affect specificity, the incremental binding energy of the P,' NH may be estimated a t -2.3 kcal/mol. 4b [ tO-CH,-C(O)-OMe] versus 15 [ tO-CH,-C(O)CH,-CH,] has an energetic penalty of AAGob = 0.4 kcal/mol for placing an oxygen a t the P i NH position compared with a -CH2-. The difference then in free energy between oxygen [4b [to-CH,-C(0)-OMe]] and N H [13 [ tO-CH,-C(O)NH,]] at that position would be AAGob = 2.7 kcal/mol. Thus, a t the P,' NH position, the specificity is NH >> CH2 = 0. This observation is in contrast to the conclusions of Merz and Kollman (1989). In free-energy perturbation simulations of transition-state analogue inhibitors of thermolysin, they concluded that the hydrogen-bonded NH of a particular inhibitor could be substituted with a CH, with only a small increase in K i , while an oxygen in the same position would decrease binding by 3.6 kcal/mol; that is, specificity is NH i= CH2 >> 0. These conclusions were supported by the Ki's determined for the inhibitors under question (Merz & Kollman, 1989). Clearly, different conditions obtain here. The energetics of interactions a t hydrogen-bonding sites on the enzyme are too complex to admit of simple qualitative explanations and must be probed empirically or computationally. Further evidence for a hydrogen-bond acceptor for the P i N H is the very poor kcat/KMI 1.1 M-'.s-' for 5 [tO-CH,COOH] where AAGobs1 4.2 kcal/mol versus 13 [ tO-CH2C(0)-NH,]. Removal of the -NH2 accounts for 2.3 kcal/mol of this, with the other 11.9 kcal/mol being accounted for by the unfavorable interaction of the carboxylate with the hy-

Papain-Substrate Interaction Energies drogen-bond acceptor. Finally, a carboxylate shifted by one methylene further away from the catalytic site interacts favorably with the enzyme, indicating the position dependence of the binding energy penalty of the carboxylate [Nbenzoyl-GlytS-CH2-CH3 versus N-benzoyl-GlytS-(CH,),COOH, AAGob, = 0.5 kcal/mol (Lowe & Williams, 1965)]. The evidence for the presence of hydrogen bonding partners for the PI’ C=O and the P,’ N H is in contrast with the observations of Brubacher and Bender (1 966), who found that the presence of this amide functionality did not increase the second-order rate constants for added nucleophiles in the deacylation of trans-cinnamoyl-papain. The work of Alecio et al. ( 1 974), however, tends to support the role of the P,’-Pi amide in the hydrolysis of amides. Brubacher and Bender (1966) found that there was no decrease in the second-order rate constants between L-tryptophanamide and L-tryptamine or methyl L-tryptophanate. Additionally, the similarity of specificity of glycinamide and aminoacetonitrile implies that the -NH2 of glycinamide does not form a hydrogen bond. There are two different principles in effect with the tryptophan-based and the glycine-based nucleophiles. The large difference between the second-order rate constants for Ltryptophanamide and glycinamide in spite of their very similar inherent nucleophilicites indicates that there is binding of the tryptophan to the enzyme in a quasi-equilibrium step before nucleophilic attack on the acyl-enzyme by the amine group. The lack of effect of the PI’-P,’ amide indicates that the binding of the tryptophan side chain dominates the association of nucleophile with enzyme and that the amide (or methyl ester) is in a different orientation than the amide of 13 [TOCH2-C(0)-NH,]. The difference in position need not be large given the strong length dependence and directionality of hydrogen bonds [particularly hydrogen bond donors (Baker & Hubbard, 1984)] and given the fact that a weak hydrogen bond is worse than none at all (Wilkinson et al., 1983). In the case of glycinamide and aminoacetonitrile, it is unlikely, due to entropic effects, that an adsorptive complex is formed between the enzyme and the nucleophiles in the SI’ subsite. The loss of entropy entailed in binding a small molecule to an enzyme can be estimated variously to be worth 10-1 5 kcal/mol. Clearly, the two hydrogen bonds through -C(0)NH2, even in the presence of strong dispersion energies acting on the C, of glycinamide (for which there is no evidence for or against) will not lead to appreciable association of the nucleophile to the enzyme in a quasi-equilibrium step before nucleophilic attack. The observed second-order rate constants represent, then, the attack of the unbound nucleophile onto the acyl-enzyme. P2and P I Subsites. ( A ) P2 NH Hydrogen Bonding. The improvement to k J K M by -3.0 kcal/mol between 1 (NH, Phe, NH)6 and 7 (H, Phe, N H ) indicates that the substrate’s P3-P2 amide hydrogen bonds with the enzyme. Exchanging an ester for the amide group decreases specificity by AAGobs = 2.7 kcal/mol [2 (0,Phe, N H ) versus 1 (NH, Phe, NH)], suggesting that it is the amide N H that is primarily responsible for the observed binding energy. If the P, C=O were to hydrogen-bond to the enzyme, then a more negative AAGobsbetween 2 (0,Phe, N H ) and 7 (H, Phe, N H ) would be expected than the weakly favorable -0.3 In this section, to facilitate the discussion of substrates involving the S2 and SI subsites. the presence or absence of various groups will be denoted in parentheses after the compound number as follows: the atoms or groups occupying the positions for the P2 NH, P2 side chain, and P, N H are shown. Thus, MwPhe-O-CH,-C(O)tOMe is 4a (NH, Phe, 0). A proton is considcrcd thc equivalent of a Gly “side chain“.

Biochemistry, Vol. 30, No. 5, 1991 1399 -0.921

02

0.123

E93 0.079

H3 0.324 -0,879

02

El 0.347

FIGURE 2: Atomic labeling for (top) N-methylacetamide and (bottom) methyl N-methylcarbamate, including atomic charges.

kcal/mol. At least part of the 0.3 kcal/mol is accounted for by differences in conformational entropy. The presence of the acetyl group of 2 hinders internal rotations of those bonds analogous to xI and iP of a Phe residue, meaning that less entropy must be lost on forming the enzyme-substrate complex. Also, in the X-ray crystallographic structures of achloroketone-inhibited papains where the inhibitor extends as far as the P5 subsite, the P3 C=O is solvent-exposed and not in contact with the enzyme (Drenth et al., 1976). The use of a methyl carbamate group in 3 (NH, Phe, N H ) in the place of an acetamide in 1 (NH, Phe, N H ) leads to a loss of specificity by AAGobs = 0.7 kcal/mol. Ab initio selfconsistent field (SCF) calculations of the charge distributions of carbamate- and amide-containing model compounds were performed by using distributed multipoles analysis (Stone, 1981) (see Appendix). While the partial positive charges on the hydrogen-bonding protons of the carbamate and amide are nearly equal in vacuo (0.347e and 0.324e, respectively; Figure 2), hydrogen bonds from the carbamate would be less favorable due to the partial negative charge of -0.626e on the nearby sp3 oxygen. Over 7630 grid points surrounding these N H groups, the average difference in electrostatic potential in vacuo is -9.4 kcal/mol, favoring the amide as a hydrogen-bond donor. The electrostatic potential in vacuo is not related to hydrogen-bonding free energy in solution in any simple manner, but it does clearly indicate that amides are better hydrogen-bond donors than carbamates. The AAGobson addition of a P3-P2 carbamate, that is, the addition of a hydrogen bond at P2 N H , depends on the presence of other enzymesubstrate interactions. Specifically, the presence of a PI N H makes the added P2 N H more favorable than when there is an ester at P2-P,; i.e., 3 (NH, Phe, NH) versus 7 (H, Phe, NH), LUGobs = -2.3 kcal/mol, and 4a (NH, Phe, 0) versus 8a (H, Phe, 0),AAGOb, = -1.0 kcal/mol. The same effect is observed in the presence or absence of the P2 Phe side chain; 3 (NH, Phe, N H ) versus 7 (H, Phe, NH), AAGob, = -2.3 kcal/mol, and 11 (NH, Gly, N H ) versus 12 (H, GIy, NH), AAGobs = -1.6 kcal/mol. ( B ) P2 Side Chain. k,,,/KM increases as the S2 binding subsite is progressively occupied with hydrophobic interactions from a number of different side chains (unpublished results). On addition of a benzyl group to 11 (NH, Gly, NH) to give 3 (NH, Phe, NH), AAGob, = 4.0 kcal/mol. The higher conformational entropy of Gly over Phe may account for part of this difference. The calculations of Brant et al. (1967) in-

1400 Biochemistry, Vol. 30, No. 5, 1991

dicated that the conformational entropy of a Gly residue about the @ and \k angles is only slightly higher than that of an amino acid with a non-&branched side chain, with an upper limit on the order of 1.5 cal.K-’.mol-l. If Phe’s lower conformational entropy is directly converted into binding energy, specificity would increase by a maximum of 0.4 kcal/mol with respect to a substrate with a P2Gly. As is the case with the hydrogen-bonding interactions, the effect of side-chain binding in S2is interdependent with other substrate-enzyme interactions. The added side chain has the greatest effect when the P2 and PI NH’s are present, as witnessed by the difference for 12 (H, Gly, NH) versus 7 (H, Phe, N H ) (PACobs= 3.3 kcal/mol) and for 11 (NH, Gly, N H ) versus 3 (NH, Phe, N H ) (AAGOb,= 4.0 kcal/mol). (0P I Hydrogen Bonding. The presence of the hydrogen bond to the PI N H was demonstrated and its strength evaluated as for the P, NH. Substitution of a P2-PI ester for the amide demonstrates that the PI N H is hydrogen-bonded to the enzyme with AAGobs = -1.2 to -2.6 kcal/mol: 3 (NH, Phe, NH) versus 4a (NH, Phe, 0),AAGobs = -2.6 kcal/mol, 7 (H, Phe, NH) versus Sa (H, Phe, O ) , Ahcobs= -1.3 kcal/mol, and 18 (-, benzoyl, N H ) versus 19 (-, benzoyl, 0), APGob, = -1.2 kcal/mol. The X-ray structures of a-chloroketone and stefin B inhibited papains (Drenth et al., 1976, Stubbs et al. 1990) indicate that the Asp 158 backbone carbonyl is the probable hydrogen-bond acceptor. Once again, the AAGobsof the PI N H depends on other enzyme-substrate interactions, with the presence of the P, N H and the P, Phe side chain cooperating to increase the effect of the PI NH. Ester for Amide Substitution. The AAGob’s for the P, (-2.7 kcal/mol) and P I NH’s (-2.6 kcal/mol) are based, at least in part, on the substitution of amide functionalities with esters at the site of the N H in question and therefore represent specificity energies. Initially, one might argue that these free energies represent a specificity against the sp3 oxygen rather thanfor the amide NH. However, because an sp3 oxygen has a smaller van der Waals radius than an N H in the hydrogen-bonding direction, selection against an oxygen would have to be electrostatic rather than steric. Any putative negative charge on the enzyme that could interact unfavorably with an ester’s sp3 oxygen would necessarily interact favorably with an amide N H ; therefore, any discrimination by the enzyme between esters and amides will include a considerable favorable free-energy contribution from the amide NH. For the P,’ N H , in contrast to the P I and P2 NH’s, the binding energy of -2.3 kcal/mol was determined independently of the ester and represents a true binding energy. The penalty to kcat/KMon placing an ester’s alcoholic oxygen adjacent to the enzyme’s hydrogen-bond acceptor in the S i subsite is approximately 0.4 kcal/mol as determined by the comparison of 4b with 15. It is not possible to directly predict the energetic penalty for an oxygen in the position of the PI N H and P, NH; however, the AAGobs L 5.2 kcal/mol for the addition of a methyl carbamate to 17 to give 4b argues for a strong hydrogen bonding of the P I NH. There is no hydrogen-bond donor in the position that the sp3 oxygen would be expected to occupy, and the terminal methyl group is positioned adjacent to Gly 66 NH. A binding energy of better than -5.2 kcal/mol is too large to have been contributed by the C=O alone. Even assuming 1 kcal/mol in favorable dispersion energy being contributed by the H3C-O-, a contribution on the order of 2 kcal/mol from the NH is required. This compares favorably with the specificity energy of -2.6 kcal/mol between ester and amide functionalities at this site and therefore represents the

Berti et al. favorable binding energy of the NH. In the case of the P, NH, it is not possible formally to estimate the incremental binding energy. However, given the known proximity of a hydrogen-bond acceptor (Gly 66 C 4 ) in the X-ray structures of a-chloroketone-inhibitedpapains (Drenth et al., 1976), it is likely that, of the AAGob = 2.7 kcal/mol for an ester versus an amide at P,-P, (2 versus l), the incremental binding energy of the P, N H will be on the order of 2 kcal/mol. Interdependence of Interaction Energies. The observed interdependence, or cooperativity, of effects at the P, NH, the P, side chain, and the P, N H are typical of enzyme-substrate interactions. While certain processes are recognized as being highly cooperative, such as protein folding (Creighton, 1983) and the association of myosin subunits (Erickson, 1989), and while the importance of cooperativity has been noted in theoretical discussions of catalysis (Jencks, 1975, 1981; Page & Jencks, 1971), there is a dearth of empirical evidence, and clear manifestations of cooperativity have not been recognized as such (Kowlessur et al., 1989). Our observation of increasingly favorable binding energies with better substrates is, to our knowledge, the first report of a gradation of cooperative effects. When a favorable interaction, for example a hydrogen bond, is added to an enzyme-substrate complex, there is a loss of internal entropy of the complex. This includes contributions from the restriction of vibrational and overall rotational motions of the substrate and enzyme relative to each other, as well as restriction of dihedral bond rotations. This loss of entropy causes, then, an increase in the free energy of the enzyme-substrate complex (E-S) and the transition state (ESI). The magnitude of the free energy increase depends on the tightness of binding of the substrate before the addition of the hydrogen bond. Thus the incremental specificity energy of the hydrogen bond is offset to a certain extent by this higher free energy of the enzyme-substrate complexes (Le., E-S and ES’). A “loose” complex where there is considerable freedom of motion of the substrate in the active site will undergo a greater entropy loss than a “tight” complex. The observable effect of this condition is that, on the addition of a hydrogen bond of a given incremental specificity energy to a substrate, the observed change in free energy (AAGob) depends on the presence of other interactions and will be greater when it is added to a good versus a poor substrate. This effect is illustrated by comparing the addition of a PI N H to 4a (NH, Phe, 0) versus adding the same PI N H to Sa (H, Phe, 0)or 19 (-, benzoyl, 0). The AAGob’s are -2.6, -1.3, and -1.2 kcal/mol, respectively. The presence in 4a (NH, Phe, 0)of the P2 N H [and P2 Phe side chain compared with 19 (-, benzoyl, O ) ] reduces the motions of complexed substrate and enzyme molecules relative to each other, making the entropic penalty upon addition of the PI N H less than with the other two substrates. It is not necessary to specifically consider ground-state destabilization to account for the interdependence of interaction energies. Changes in the second-order rate constant kcat/KMreflect changes in the overall binding energy that is available, which may be expressed as changes in either k,,, or KM.. The concept of ground-state destabilization is concerned only with the partitioning of binding energy between formation of the noncovalent enzyme-substrate complex and decreasing the activation energy of the reaction, not the total amount of binding energy available. Alternately phrased, because the entropy of the enzymesubstrate complex is higher for poor substrates, the restriction of motion necessary to form the hydrogen bond with the PI

Papain-Substrate Interaction Energies Z

FIGURE 3: Coordinate system used for investigating region around hydrogen-bonding NH’s.

N H raises the free energy of the enzyme-substrate complex (E&) and the transition state (ES*) more for poor substrates than for good. In each case, the N H group interacts equally strongly with the hydrogen-bond acceptor, but for poor substrates a larger portion of the interaction energy compensates for the higher free energy of the complex and therefore appears as a poorer AAGoh. Precisely the same effect is observed with the P2 NH and the P2 Phe side chain as with the P I NH. This principle is further illustrated by the results of Asboth et al. ( 1 988), as compared with those of Williams et al. (1 972). Substitution of the P, C=O of Z-PheGlytOEt (kcat/KM= 120 000 M - ’ d for papain) with C=S decreased specificity by 4.4 kcal/mol for papain (Asboth et al., 1988). These AAGoh’s represent the loss of the C=O hydrogen bond plus the unfavorable free energy of fitting the C=S into a site that normally accommodates the smaller C=O. For N-benzoylGlytOMe (kca,/KM = 246 M - k ’ ) , AAGobs = 1.6 kcal/mol on substituting the P, carbonyl oxygen with a sulfur atom (Williams et ai., 1972). As with the substrates reported in this study, the effect of oxygen to sulfur substitution at P2 C=O depends on other enzyme-substrate interactions. The better the substrates, the more unfavorable the resulting effect. The AAGob’s for the interactions discussed above represent the lower limits of incremental specificity (P, N H , PI N H ) or binding (P2 Phe side chain) energies. In particular, the incremental binding energy for a P2 Phe side chain is more than the -4.0 kcal/mol observed here between 3 and 11. The poorer of the two substrates in the comparison, 11, is quite poor indeed and may therefore have a much higher entropy in the E-S and ES* complexes than the corresponding species for 3. With the leaving-group interactions, the AAGobs’sare based on relatively poor substrates and therefore may also significantly underestimate the incremental binding energies. The strengths of the hydrogen bonds relative to each other (PI’ C=O, AAGoh = -1 .O kcal/mol; P,’ NH, -2.3 kcal/mol) may represent differences in the entropy of hydrogen-bond formation. Upon formation of the hydrogen bond to the PI’ C=O, the movement of the P,’ N H becomes very restricted due to the planar geometry of the amide, making the entropic cost of forming the P,’ N H hydrogen bond relatively smaller. Application of this observed cooperativity of energies of interaction is in the design of synthetic inhibitors of cysteine proteases, or indeed, in the manipulation of intermolecular binding in general. The highest possible specificity of an inhibitor for one protease over another will be achieved by utilizing all the bonding interactions common to the two enzymes, maximizing the effectiveness of those interactions that are different. For example, considering two hypothetical cysteine proteases that share the same peptide backbone hydrogen bonding as observed here but have different specificities for the P2 side chain, the use of all the common backbone

Biochemistry, Vol. 30, No. 5, 1991

1401

interactions in an inhibitor will allow the incremental binding energy of the P, side-chain interactions to be expressed as specificity between the two proteases and not be “lost” in entropy compensation. The hydrogen bonds to the substrate’s peptide backbone at P2 NH, P2 C=O, and PI N H all involve backbone atoms of the papain molecule: Gly 66 C 4 , Gly 66 NH, and Asp 158 C=O, respectively. As such, these hydrogen-bonding interactions are expected to be conserved through most or all cysteine proteases. It would therefore be useful to include these interactions in any synthetic inhibitors of cysteine proteases in order to maximize selectivity. In conclusion, we have shown that the active-site cleft of papain has hydrogen-bonding partners for the peptide backbone of substrates extending at least from the S2 to the S2’ subsites. The lower limits of the incremental specificity energies for the P, N H and P I NH’s are, respectively, -2.7 and -2.6 kcal/mol versus esters at these sites. The incremental binding energies are estimated to be on the order of 2 kcal/mol. The AAGo,’s for P2 Phe side chain (-4.0 kcal/mol), PI’ C=O (-1 .O kcal/mol), and P,’ N H (-2.3 kcal/mol) probably all underestimate the incremental binding energy at these points. The observed interdependence of binding energies at each site on the acyl side of the catalytic site is characteristic of enzymesubstrate interactions and is largely due to a decreased entropic penalty with successive additions of favorable interactions. The use of all the backbone hydrogen bonding interactions observed here will aid in producing synthetic inhibitors of cysteine proteases with the greatest possible selectivity. ACKNOWLEDGMENTS We gratefully acknowledge Dr. Ron Angus for the gift of several substrates. APPENDIX It is well-known that the electrostatic potential can describe the hydrogen-bonded geometries in small van der Waals complexes (Buckingham & Fowler, 1985). We have compared the electrostatic potential around N H for both amide- and carbamate-containing molecules as a means to assess their tendency to form hydrogen bonds. The electrostatic potential was calculated by using a distributed multipoles analysis (Stone, 1981) of the SCF wave function, calculated with a 3-21G basis set by using the a b initio program suite CADPAC (Amos, 1984). The atomic dipoles, quadrupoles, and higher moments explicity represent the nonspherical features in the atomic charge distribution, such as the lone-pair density. The molecules were superimposed in such a way that the N H groups had identical coordinates, thus facilitating the comparison of the electrostatic potential. Figure 2 illustrates both molecules with the atomic labeling, while Figure 3 illustrates the coordinate system chosen to describe the vicinity of the N H bond. Some geometrical parameters from the work of Taylor et al. (1983) on the geometry of the NH-.O=C hydrogen bond were used. The electrostatic potential was studied in the region of space around the N H bond, namely 4 € 2 7 O and 0 € 360° (both angles were stepped in 3O increments). In addition, the distance from the nitrogen to the observation point was varied from 2.6 to 3.2 in 0.1-A steps. This gave 7630 grid points in the vicinity of NH. The electrostatic potential from the atomic multipole moments up to rank 4’ was calculated by using ORIENT (Price & Stone, 1987).



A complete list of these atomic moments is available from the authors upon request.

1402 Biochemistry, Vol. 30, No. 5, 1991 The difference between the electrostatic potential of amide and carbamate was calculated by using AV(i) = V(i)carb - V(?)amide where i: is within the region described above. The results obtained are Ab‘(?),, = -9.4 kcal/mol over the 7630 grid points. This clearly shows in a quantitative manner that the electrostatic potential of carbamate is systematically smaller than that of amide in the region around the potentially hydrogen bonding NH. This is in accord with the experimental observation that carbamates are worse hydrogen-bond donors than amides. Registry No. 1, 7625-55-0; 2, 131041-20-8; 3, 101653-92-3; 4, 131010-72-5; 5, 131010-73-6; 6, 41844-91-1; 8, 131010-74-7; 9, 131010-75-8; 10, 501-52-0; 11, 131010-76-9; 12, 1906-82-7; 13, 131010-77-0; 14, 131010-78-1; 15, 131010-79-2; 16,41844-71-7; 17, 60397-80-0; 18, 1205-08-9; 19, 29747-05-5; AcPhe, 201 8-61-3; GlyOMe. 61 6-34-2; papain, 9001 -73-4; bromoacetate, 68-1 0-0; iodoacetamide, 144-48-9; 1-hydroxypropanone, 1 16-09-6; 1-hydroxy2-butanonc, 5077-67-8.

REFERENCES Alecio, M. R., Dann, M. L., & Lowe, G. (1974) Biochem. J . 141, 495-501. Amos, R. D. (1984) CADPAC:The Cambridge Analytical Derivatives Package, Publication CCPl/84/4 91984, SERC Daresbury Laboratory, Daresbury, Warrington, U.K. Angus, R. H., Carey, P. R., & Storer, A. C. (1986) Biochemistry 25, 3304-33 10. Asboth, B., Majer, Z., & Polgar, L. (1988) FEBS Lett. 233, 339-341. Baker, E. N., & Hubbard, R. E. (1984) Prog. Biophys. Mol. Biol. 44, 97-179. Berger, A., & Schechter, 1. (1970) Phiios. Trans. R . Soc. London B 257, 249-264. Berti, P. J. ( 1 990) M.Sc. Thesis, University of Ottawa, Ottawa, Ontario, Canada. Bjorck, L., Akesson, P., Bohus, M., Trojnar, J., Abrahamson, M., Olafsson, I., & Grubb, A. (1 989) Nature 337, 385-386. Brant, D. A., Miller, W. G., & Flory, P. J. (1967) J. Mol. Biol. 23, 41-65. Brocklehurst, K., Willenbrock, F., & Salih, E. (1987) in Hydrolytic enzymes (Neuberger, A., & Brocklehurst, K., Eds.) pp 39-1 58, Elsevier Biomedical Press, Amsterdam. Brubacher, L. J., & Bender, M. L. (1966) J. Am. Chem. SOC. 88, 5871-5879. Buckingham, A. D., & Fowler, P. W. (1985) Can. J . Chem. 63, 201 8-2025. Carey, P. R., Angus, R. H., Lee, H., & Storer, A. C. (1984) J . Biol. Chem. 259, 14357-1 4360. Cohen, S . G., & Weinstein, S . Y. (1964) J. Am. Chem. SOC. 86, 5326-5330. Coombs, G . H. (1982) Parasitology 84, 149-155. Cornish-Bowden, A. (1975) Biochem. J. 149, 305-312. Cornish-Bowden, A. (1976) in Principles of enzyme kinetics, Butterworths, London. Creighton, T. E. ( 1 983) in Proteins, structures and molecular properties, W. H. Freeman and Company, New York.

Berti et al. Demuth, H.-U. (1990) J . Enzyme Znhib. 3, 249-278. Drenth, J., Kalk, K. H., & Swen, H. M. (1976) Biochemistry 15, 3731-3738. Dresden, M. H., Payne, D. C., & Basch, P. F. (1982) Mol. Biochem. Parasitol. 6 , 203-208. Erickson, H. P. (1989) J. Mol. Biol. 206, 465-474. Fersht, A. R. (1988) Biochemistry 27, 1577-1580. Finklestein, A. V., & Janin, J. (1989) Protein Eng. 3, 1-3. Ingles, D. W., & Knowles, J. R. (1968) Biochem. J. 108, 56 1-569. IUPAC-IUB Joint Commission on Biochemical Nomenclature (1985) J. Biol. Chem. 260, 14-42. Jencks, W. P. (1975) Ado. Enzymol. Relat. Areas Mol. Biol. 43, 219-410. Jencks, W. P. (1981) Proc. Natl. Acad. Sci. U.S.A. 78, 4046-4050. Kowlessur, D., Topham, C. M., Thomas, E. W., O’Driscoll, M., Templeton, W., & Brocklehurst, K. (1989) Biochem. J. 258, 155-764. Lowe, G., & Williams, A. (1965) Biochem. J . 96, 199-204. Merz, K. M., Jr., & Kollman, P. A. (1 989) J. Am. Chem. SOC. I 1 I , 5649-5658. Nefkens, G. H. L., & Nivard, R. J. F. (1965) Recl. Trau. Chim. Pays-Bas 84, 13 15-1 3 18. Page, M. I., & Jencks, W. P. (1971) Proc. Natl. Acad. Sci. U.S.A. 68, 1678-1683. Polgar, L. (1987) in Hydrolytic enzymes (Neuberger, A,, & Brocklehurst, K., Eds.) pp 159-200, Elsevier, Amsterdam. Price, S . L., & Stone, A. J. (1987) J . Chem. Phys. 86, 2859-2868. Rich, D. H. (1986) in Proteinase inhibitors (Barrett, A. J., & Salvesen, G., Eds.) pp 153-178, Elsevier, Amsterdam. Schechter, I., & Berger, A. (1967) Biochem. Biophys. Res. Commun. 27, 157-162. Schechter, I., & Berger, A. (1968) Biochem. Biophys. Res. Commun. 32, 898-902. Shah, H. N., & Gharbia, S . E. (1989) FEMS Microbiol. Lett. 61, 213-218. Sluyterman, L. A. A. E., Wijdenes, J. (1970) Biochim. Biophys. Acta 200, 593-595. Stone, A. J. (1981) Chem. Phys. Lett. 83, 233-239. Storer, A. C., & Carey, P. R. (1985) Biochemistry 24, 6808-68 18. Storer, A. C., Lee, H., & Carey, P. R. (1983) Biochemistry 22,4789-4796. Storer, A. C., Angus, R. H., & Carey, P. R. (1988) Biochemistry 27, 264-268. Stubbs, M. T., Laber, B., Bode, W., Huber, R., Jerala, R., Lenarcic, B., & Turk, V. (1990) EMBO J . 9 , 1939-1947. Taylor, R., Kennard, O., & Vershichel, J. (1983) J . Am. Chem. SOC.105, 5761-5766. Whitaker, J. R., & Bender, M. L. (1965) J. Am. Chem. SOC. 87, 2728-2737. Wilkinson, A. J., Fersht, A. R., Blow, D. M., & Winter, G. (1983) Biochemistry 22, 3581-3586. Williams, A., Lucas, E. C., Rimmer, A. R., & Hawkins, H. C. (1972) J. Chem. SOC.,Perkin Trans. 1 , 627-633.